IBM has announced significant updates to Watson Natural Language Understanding that it said will enable organisations to better gauge customer sentiment, as well as achieve faster and more accurate processing of documents such as PDFs at scale.
It amounts to the commercial debut of key IBM Natural Language Processing (NLP) capabilities emerging from its ‘Project Debator’, which it says is the only AI system able to engage humans in complex debates.
New features will be added progressively throughout this year, the company said.
Among the key capabilities now being brought to market with Watson is the ability to understand idioms and colloquialisms such as ‘hardly helpful’ or ‘hot under the collar’, which have tended to confuse AI systems to date.
Such phrases are known as ‘sentiment shifters’, and with advanced sentiment analysis, businesses can now make better use of this data to gain a more holistic understanding of operations.
In addition to advanced sentiment analysis, IBM is also adding new capabilities for ‘summarization’ as well as ‘clustering’.
Summarization pulls textual data from multiple sources providing users with an overview of what is being said and written about a particular topic. A previous version of this technology was deployed earlier this year to analyse over 18 million articles, blogs and bios as part of a large-scale data project undertaken for the Grammys music awards. Data was used to build a live stream for the red carpet segment with on-demand videos and photos giving fans deeper context on leading topics.
New topic clustering techniques will enable users to create meaningful “topics” of related information for analysis. IBM said the technique will be integrated into Watson Discovery later this year, and will also allow subject matter experts to customise and tweak topics to reflect the language of different businesses or industries such as insurance, healthcare and manufacturing.
“Language is a tool for expressing thought and opinion, as much as it is a tool for information,” said Rob Thomas, general manager, IBM Data and AI.
“This is why we’re harvesting technology from Project Debater and integrating it into Watson – to enable businesses to capture, analyse, and understand more from human language and start to transform how they utilise intellectual capital codified in data.”
IBM already has a number of high profile customers using AI-driven language processing in Australia, including financial services company, Suncorp.
Jason Leonard, partner, cognitive computing and analytics at IBM Australia told CIO the new language features announced for IBM Watson would enable significant improvements in sentiment analysis across key areas of their operations.
Already Suncorp has reported encouraging results in its insurance claims division, using a Watson-based system able to understand simple yet important distinctions such as “the car crashed into the back of me” and “I crashed into the back of the car”.
Lisa Harrison, Suncorp’s chief customer and digital officer said the company has been using Watson natural language to improve customer experiences since 2017, as well as simplifying processes freeing staff to work on more meaningful tasks.
“Each year we receive more than 500,000 motor insurance claims and implementing this technology has significantly reduced the time taken for customers to complete a claim, with many doing so in as little as five minutes,” she said.
“IBM Watson is smart, fast and reliable, and we are excited to seeing how the new NLP features will enable us to deliver more great outcomes for our customers.”
Other notable Australian customers include energy giant Woodside, which is using Watson language processing to collect and analyse language data for improving worker safety, while Sydney-based media company Oovoo is using Watson language processing to vastly speed up the process of selecting videos to accompany articles online.
Internationally, ESPN Fantasy Football has deployed Watson Discovery and Watson Knowledge Studio to crunch millions of football data sources each day of the season, providing millions of fantasy football players with real-time insights.
Watson uses natural language processing to discern the tone and sentiment of multiple data sources including news articles, forums, blogs, projections, rankings, tweets and podcasts that might cover everything from locker-room chatter to injury analysis.
ESPN Fantasy Football then “surfaces” these discoveries in so-called player cards that note the “boom and bust” potential of players, while also informing the ‘Player Buzz” section which summarises positive or negative commentary about a player.
A natural extension of IBM’s infamous foray into the Jeopardy game show world, Debator has also proved itself a capable adversary with a recent debate being only narrowly won by a human according to a live audience, which voted the ‘machine’ as a better source of knowledge.
It’s an interesting development for industries and professions where language and communication are prominent.
“We can see this technology playing an important role in the legal profession where machines could be used to augment the activities of lawyers actually presenting in court,” said IBM’s Leonard.